FR3037176A1 - DEVICE FOR DETECTING THE PRESENCE OF PERSON ON THE GROUND BY ANALYSIS OF A FLOW OF DEPTH MAPS - Google Patents

Abstract

The invention relates to a method and a device for detecting the presence of a person on the ground and for transmitting an alert. It consists of a depth card acquisition module, a processing module and a communication module. The processing module analyzes the depth maps in order to determine, in an initialization phase, the ground plane, then, then in a phase of permanent operation, to detect the presence of new objects and to determine by analysis of geometric and temporal criteria. there is presence on the ground. The device according to the invention is particularly intended for the detection of the presence on the ground of elderly people or dependent on their place of life.

Description

The present invention relates to a ground presence detection method by analyzing a flow of depth maps from a 3D scanning optical module. PRIOR ART Today it is very difficult to detect falls of old people in a reasonable time, especially at night. This is all the more damaging since, after several minutes on the ground, additional sequelae may occur, which may even be life-threatening. The fall is the number one killer of over 65 years old. To give the warning in case of a fall, the oldest solution is in the form of a medallion or a bracelet of voluntary call allowing the carrier to prevent in the event of problem such as a fall, but it may forget or refuse to wear it and, in case of discomfort, be unable to operate it. Other devices automatically detect the fall by analyzing sensor signals placed in a watch or a patch glued to the torso. Again, the person must agree to wear them because they are stigmatizing. When they are worn, they can also cause false alarms (during a shock of the hand for example), and detect with difficulty the soft falls (when the person slips to the ground while holding a piece of furniture or a wall, during a malaise for example). Other devices include pressure and vibration sensing mats to be placed around the bed, but the area covered is limited. More recently, sensor devices placed under the ground allow detection over the entire surface of the living place of the person, including the bathroom for some technologies, but these devices require heavy work and are very expensive.

Finally, there are solutions based on the use of video cameras associated with image processing software to analyze the horizontal and vertical speeds related to the movements of the person and determine if they correspond to a fall. This analysis requires great computing power to execute complex algorithms. In addition, these solutions hardly detect some slow drops that generate very low vertical speeds. It can be added that the use of video cameras involves the acquisition of images in which it is possible to identify people to the detriment of privacy. Finally, many systems require transmitting an image to a computer or smartphone to confirm the alarm. As a result, many people consider the permanent acquisition and transfer of images to identify individuals as intrusions into their privacy. There is therefore a real need for a method and a device for reliable fall detection, low cost and simple to implement, having nothing to wear by the person, and respecting his privacy with a guarantee of anonymity.

One solution is the use of shape sensors (or depth maps or 3D cameras) that measure the distances of objects from the scene to the sensor regardless of the color or texture of said objects. Compared to video, this type of sensor is insensitive to light variations and colors, which simplifies the analysis and provides calibrated information: the dimensions of the objects are the same regardless of the distance to the sensor. US Pat. No. 8,427,324 B2 discloses a method and apparatus for detecting a fall of a person based on distance image analysis. This method consists of: generating an initial image of the distances; identify a reference plane present in the field of view of the sensor; determining the representative regions of the objects located above the reference plane in the field of view; generate a new image of the distances corresponding to the field of vision after a determined period of time generally of 0.1s; compare the new image to the initial image to determine if the new image of the distances includes a new object placed above the reference plane; the new object is considered to have fallen if its height is less than a given height and if its volume and / or area is greater than a determined value. This method and device have the following drawbacks in particular: - The defects of the sensors are not taken into account. In particular, problems of acquisition noise, interference from other sensors or the environment, or non-reconstructed areas of the image (because of absorption or diffraction phenomena of the rays). infrared, for example) are not examined. The initial image is not updated, which, combined with the inhibition of the areas containing objects during initialization, may result in an undetected drop if it occurs in the area in question. ; - No solution is proposed to remove false alarms related to displaced objects that meet the dual condition of having a height less than the threshold height determined and a volume and / or a surface greater than a determined value; 10 - The fall will not be detected if the person comes into contact with an object moved higher than the detection threshold, such as a door. BRIEF DESCRIPTION OF THE INVENTION The present invention seeks to overcome one or more of the disadvantages of the prior art. For this purpose, the present invention relates to a method for detecting the presence of a person on the ground by analyzing a stream of depth maps acquired by a depth map acquisition module, characterized in that the method comprises the following steps: - The acquisition of N (at least one) depth maps corresponding to the field of view of the depth map acquisition module, - Calculation of a reference depth map by integrating the N maps acquired at the previous step, with elimination of aberrations due to the acquisition module, 25 - Calculation of the ground plane in the scene, - Determination of a world reference having the ground plane as zero altitude, - Acquisition of a depth map at a given rate, - Detection of the objects by comparison between the acquired map and the reference map, 3037176 4 - Analysis of the detected objects, by estimation of the height of their high point, of the height of the the center of gravity in the depth map and their surface projected on the ground, - triggering an alarm if at least one detected object verifies that the height of its high point is lower than a given height, that the height of its center of gravity in the depth map is less than a given second height and its projected area on the ground is greater than a given surface, for a time greater than a given duration, - Refreshing the reference map to account for objects displaced, when a given feedback criterion is reached. The invention also relates to a device for detecting the presence on the ground of a person by analyzing a flow of depth maps, characterized in that the detection device comprises: a means for acquiring N (at least one) depth maps 15 corresponding to the field of view of the depth map acquisition module, - means for calculating a reference depth map by integrating the N maps acquired in the previous step, with elimination of the aberrations due to the means of acquisition, 20 - means for calculating the ground plane in the scene, - means for determining a world reference frame having the ground plane as zero altitude, - means for acquiring a depth map at a given rate, - means for detecting objects by comparison between the acquired map and the reference map, - means for analyzing the objects detected, by estimating the height of their high point, of the height of their center of severity in the depth map and their projected surface on the ground, - A means of triggering an alarm if at least one detected object 30 verifies that the height of its high point is lower than a given height, that the height of its center of gravity in the depth map is less than a given second height and that its projected area on the ground is greater than a given surface, for a time greater than a given duration, - means for updating the reference map to account for moved objects when a given feedback criterion is met.

According to one embodiment, the object detection step is performed only when the temperature information provided by the thermal image acquisition module includes temperature information indicating the presence of at least one human being. . According to one embodiment, the step of analyzing the detected objects comprises the temperature reading 10 from an image acquired by a thermal image acquisition module which observes the same scene as the acquisition module of the images. maps of depths and retains only objects whose temperature is characteristic of that of a human being. According to one embodiment, the step of analyzing the detected objects comprises a criterion 15 for deleting objects whose movement is mainly horizontal with respect to the previously determined ground plane. The invention also relates to a computer program stored on an information carrier, said program comprising the instructions adapted to implement the methods described above, when it is executed by a computer system or a microprocessor. The invention also relates to storage means comprising such a computer program. According to its material aspects, the present invention relates to a ground presence detector comprising a depth map acquisition module, a depth map processing module and a communication module.

According to another of its material aspects, the present invention relates to the use of a thermal image acquisition module. According to another of its aspects, the present invention relates to a system comprising at least two ground presence detectors for improving the coverage of the surface of the part to be secured.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic representation of the device according to one embodiment of the invention; FIG. 2 is a representation of a ground presence detection method according to one embodiment; of the invention, as implemented by the device of FIG. 1. DETAILED DESCRIPTION OF AN EMBODIMENT OF THE INVENTION With reference to FIG. 1, a device (101) for detecting persons (106) at ground 10 includes a depth map acquisition module (102), a depth map processing module (103), a communication module (104), and an optional thermal image acquisition module (110) . The depth card acquisition module (102) or 3D camera may be based on various technologies such as flight time, triangulation with structured light projection, or stereovision. The depth card acquisition module must operate in the infrared spectrum, in order to operate under all illumination conditions, and not to produce visual disturbances. The depth card acquisition module (102) may be wall or ceiling mounted so as to observe the largest possible volume portion of the equipped part, while having the ground in its field of view. Typically, the acquisition module is placed on the ceiling plunging from 35 degrees to 45 degrees depending on the geometry of the fitted room. The depth card processing module (103) may be a remote or integrated computer station.

The communication module may be an all-or-nothing output, or a wired or non-wired connection respecting the standards and protocols specific to the location of the system. According to a first embodiment, the operating mode of the device is as follows.

The depth card acquisition module (102) acquires in real time a stream of depth maps, which are images each pixel of which carries the measurement of the distance to the reference plane of the point sensor of the scene seen at this pixel. The time interval separating the acquisition of two consecutive depth maps is typically of the order of 33 milliseconds. With reference to FIGS. 1 and 2, in depth map acquisition step (201), the acquisition module (102) acquires N (at least one) depth map. These N acquired depth maps are integrated in the step (202) of calculation of the reference map, to form a map of reference depths. If pixels of the reference depth map 10 are zero, that is to say that the acquisition module could not determine a value, the reference depth map is corrected by interpolation of the pixels to 0. Typically, a bilinear interpolation can be used to make this correction. In step (203) of calculating the ground plane, the depth map processing module (103) automatically determines the ground plane equation visible in the reference depth map, and determines at step (204) Determination of the world reference frame a world reference having the ground as zero altitude. The depth map processing module (103) compares each new map acquired in the depth map acquisition step (205) with the reference depth map computed at the step (202) for calculating the depth map. reference map for determining in the object detection step (206) the objects reported in the observed scene relative to the scene of the reference depth map. The detection image is segmented to define regions of interest. For each of the determined regions of interest, the depth map processing module (103) determines in the analysis step (207) of the detected objects the height of its highest point in the three-dimensional space associated with the ground plane, the height of its center of gravity, and its projected surface on the ground. The depth map processing module (103) transmits at the ground presence detection step (208) an alarm (209) if a region of interest has a high point of height less than a first threshold S1, a center of gravity of height less than a second threshold S2, and a projected area on the ground greater than a third threshold 53, for a duration greater than a fourth threshold S4. Typically, S1 may be 1 meter, S2 50 centimeters, S3 0.3 square meters and S4 1 minute, for example. The alarm (209) may be local (light, sound) or relayed by the communication module (104) in the form of a message (105), an all-or-nothing output, an input in a database or a log file for example. This relay can be accompanied by a grayscale image representing the depth map acquired at the moment the alarm was issued, for the purpose of raising doubt, for example. For each region of interest determined at the object detection step (206) in the depth map, the depth map processing module (103) determines whether the detected object is to be scanned or not. Typically, it is a matter of determining whether the detected object is an object related to the scene or an object of the scene that has moved between the generation of the reference depth map and the acquisition of the depth map common.

The analysis of the regions of interest of the detected objects analysis step (207) can relate to their characteristics, such as the analysis of their shape, their flatness, their orientation relative to the ground, by example. The device can also verify that the object corresponding to the region of interest examined does not float above the ground (above a table, or sitting of a chair, for example). In some modes of operation, the device (101) may also incorporate object tracking mechanisms over time to determine a momentum for the region of interest, for example. The regions of interest that meet the above criteria can thus be filtered so that ground presence detection is only performed on the objects reported to the scene between the generation of the reference depth map and the acquisition of the map of current depths, and the people (106) who evolve in the observed scene. The filtering thus carried out also makes it possible to avoid concatenations of regions of interest which could have caused cases of non-detection of presence on the ground (in the case where, for example, a person finds himself on the ground in front of a high object 30 moved , such as a door for example).

Similarly, the reference depth map is regularly updated in step (211) refreshing the reference, which eliminates the detection of moved objects until their next move. The criterion of reactualization (210) can be a given schedule, a given frequency, or the absence of movement in the observed scene, for example. In another mode of operation, the step (202) for creating the reference depth map may be accompanied by a step of generating a map of the reference heights. In this case, the analysis step (207) of the objects of interest will comprise a current map height generation step, which will be compared with the map of the reference heights to determine if the objects of interest are or not above an object of the scene. Thus, the case where the person (106) is lying on an undetected low object (such as a ground mattress, for example) can be elucidated, while the step (207) for analyzing objects of interest without a map of heights would trigger an alert (209).

In another mode of operation, the depth map acquisition module (102) can be coupled with sensors of different natures. For example, a thermal image acquisition module (110) may be used in conjunction with the depth map acquisition module (102) to discriminate objects detected at the object detection step (206) in the 20 depth map. In this mode of operation, the thermal image acquisition module (110) determines in the step (206) of detecting objects in the depth map if there is temperature information that indicates the presence of least one human being in the scene observed by the device (101). If in this way, no human being is detected in the observed scene, the further step of (206) detecting the objects in the depth map as well as the step (207) of analyzing these objects are passed, so that the system goes directly to the ground presence detection step (208). In another mode of operation, the thermal image acquisition module (110) returns to the step (207) of analysis of the objects detected in the depth map the temperature of these objects and determines whether it corresponds to the temperature of a human being or the temperature of the environment. In this case, objects whose temperature does not correspond to that of a human being are deleted. The proposed sensor coupling makes it possible to discriminate the objects of the persons present in the scene, it solves the problems of false alarms related to displaced objects, as well as the problems of non-detection in the case where the person is on the ground in contact with a person. detected object such that its height is greater than the detection threshold of presence on the ground. In another mode of operation, the device (101) may be coupled with one or more other devices of the same type, observing or not the same scene as the first, at a different angle if necessary to allow the zone to be expanded. secure for example, and communicating or not with this first device. INDUSTRIAL APPLICATION The device according to the invention is particularly intended for the detection of the presence on the ground of elderly people or dependent on their place of life. However, the present invention is not limited to the embodiments described and shown, but the skilled person will be able to make any variant within his mind. For example, the present invention can be adapted to address problems of fall detection and activity measurement, problems of enumeration of persons in secure locks or in queues, or problems security, surveillance and / or access control for sensitive areas.

Claims (9)

CLAIMS1) A method of detecting the presence on the ground of a person by analyzing a stream of depth maps acquired by a depth map acquisition module, characterized in that it comprises the following steps: - acquisition (201 ) N (at least one) depth maps corresponding to the field of view of the depth map acquisition module; - calculation (202) of a reference depth map by integration of the N cards acquired in the previous step, with elimination of aberrations due to the acquisition module; - calculation (203) of the ground plane in the scene; determination (204) of a world reference having the ground plane as zero altitude; acquisition (205) of a depth map at a given rate; detection (206) of the objects by comparison between the acquired card and the reference card; - analyzing (207, 208) objects detected by estimating the height of their high point, the height of their center of gravity in the depth map and their projected surface on the ground; - triggering an alarm (208, 209) if at least one detected object verifies that the height of its high point is lower than a given height, that the height of its center of gravity in the depth map is less than a second given height and that its projected area on the ground is greater than a given surface, for a time greater than a given duration; - Refactualization (211) of the reference map to account for objects moved when a given criterion of reactualization (210) is reached. 2) Method according to claim 1, characterized in that the step of detecting the objects (206) is activated only if the image acquired by a thermal image acquisition module (110) includes a temperature information indicating the presence of at least one human being. 3) Method according to claim 1, characterized in that the step of analyzing the detected objects (207) further comprises the deletion of the objects by analysis of their thermal signature obtained from an image acquired by a a thermal image acquisition module (110) which observes the same scene as the depth map acquisition module (102). 4) Method according to any one of the preceding claims, characterized in that said step (207) comprises an estimation of the movement of the detected objects and a filtering of the objects whose movement is mainly horizontal with respect to the plane of the ground determined in the steps (203, 204), 5) A computer program product comprising program instructions adapted to implement the method of any one of the preceding claims when said program is run on a computer or microprocessor. 6) Computer-readable storage means, characterized in that they store a computer program comprising the instructions for carrying out a method according to at least one of claims 1 to 4 when said program is executed on a computer or a computer. microprocessor. 7) Device (101) for detecting the presence of a person on the ground by analyzing a flow of depth maps, characterized in that the detection device comprises: - means for acquiring N (at least one) cards depths 20 corresponding to the field of view of the depth map acquisition module; a means for calculating a reference depth map by integrating the N maps acquired in the preceding step, with elimination of the aberrations due to the acquisition means; A means for calculating the ground plane in the scene; a means for determining a world reference frame having the ground plane as zero altitude; means for acquiring a depth map at a given rate; means for detecting objects by comparison between the acquired card and the reference card; A means for analyzing the detected objects, by estimating the height of their high point, the height of their center of gravity in the depth map and their projected surface on the ground; means for triggering an alarm if at least one detected object verifies that the height of its high point is less than a given height, that the height of its center of gravity in the depth map is less than a second height given and that its surface projected on the ground is greater than a given surface, for a time greater than a given duration; a means of updating the reference map to take account of the objects moved when a given feedback criterion is reached. 8) Device (101) according to the preceding claim, characterized in that it further comprises a thermal image acquisition module (110) and a means for deleting the objects by analysis of their thermal signature obtained from an image acquired by a thermal image acquisition module (110) which observes the same scene as the depth map acquisition module (102). 9) system for detecting the presence on the ground of a person characterized in that it comprises at least two devices (101) according to one of claims 7 to 8.